Posts by Drand@techhub.social
(DIR) Post #ASPcZOJtPgTbMvadGK by Drand@techhub.social
2023-02-05T20:17:42Z
0 likes, 1 repeats
How big of a platform does Twitter offer? Now that Musk made view counts public, we (w/ @mohsenmosleh ) can find out... ANSWER: Not nearly as big as it looks! For accounts with lots of followers, only a small fraction (~1%) see median tweet 😬 (and this overcounts, as it includes RT views)
(DIR) Post #ASPcZPciZB83PbPBFA by Drand@techhub.social
2023-02-05T20:17:43Z
0 likes, 0 repeats
From a list of 300M users, @mohsenmosleh sampled 611 users, stratified on log(followers), and collected all 187k of their primary tweets since Twitter made view counts public.
(DIR) Post #ASPcZQnOCzG92zPDW4 by Drand@techhub.social
2023-02-05T20:17:44Z
0 likes, 0 repeats
Why are view counts surprisingly low for large-follower accounts? Some ideas:*Lots of those followers are fake (bots)? *Algorithm is hugely important, more so than # followers?*Lots of people have quit Twitter?Other ideas?
(DIR) Post #ASPcZRva01PAYgFGvA by Drand@techhub.social
2023-02-05T20:17:44Z
0 likes, 0 repeats
Addendum: I've been using Mastodon a lot, and have found the ratio of engagement to followers to be WAY higher here than on Twitter
(DIR) Post #ASdw9NTul2QIfETrii by Drand@techhub.social
2022-11-21T03:52:36Z
1 likes, 2 repeats
Hi Mastodon! I do research on #misinformation #fakenews #socialmedia & #polarization at MIT. For example, I've found that(i) people online are often distracted from thinking about accuracy s/t shifting attention back to accuracy increases the quality of news shared - see for example https://www.nature.com/articles/s41586-021-03344-2(ii) the wisdom of crowds can be harnessed to identify misinfo at scale - see for example https://www.science.org/doi/10.1126/sciadv.abf4393I mostly post about new research on those topics, and occasionally about #punkrock bands. For an overview of our research check out this doc: https://docs.google.com/document/u/0/d/1k2D4zVqkSHB1M9wpXtAe3UzbeE0RPpD_E2UpaPf6Lds/mobilebasic
(DIR) Post #AUl6j7yaJqrsyjrbBQ by Drand@techhub.social
2023-04-17T23:52:32Z
0 likes, 0 repeats
@garymarcus Ziv Epstein has an experimental paper on anthropomorphizing AI you might like https://www.sciencedirect.com/science/article/pii/S2589004220307070
(DIR) Post #AXNRIK9o1iKezews0O by Drand@techhub.social
2023-06-29T17:05:38Z
1 likes, 5 repeats
"Understanding and combatting #misinformation across 16 countries on six continents" out today in Nature Human Behavior! There have been a huge number of msinfo-related studies in the past 5 or so years, but almost all of them have been US/Western populations. So here we examined the #psychology of #misinformation, and the effectiveness of various #interventions, across 16 countries (total N=34k). We find: ➤Consistent #cognitive, #social & #ideological predictors of misinfo belief across countries➤Interventions (accuracy prompts, digital literacy tips, using crowdsourcing to identify misinfo) are all broadly effective across countriesGive it a read and let me know what you think! https://rdcu.be/dfGtT
(DIR) Post #AcX2K6cmknIxl6K3Q8 by Drand@techhub.social
2023-06-20T18:15:55Z
1 likes, 0 repeats
🚨Out in PNAS!🚨https://www.pnas.org/eprint/AWU8WZ9GVGPPGP9FJT3D/full#Political #microtargeting has become a topic of great concern (eg Cambridge Analytica) But does it actually WORK? To find out, we ran a large-scale experiment to quantify the persuasive advantage of political microtargeting for US issue advocacy campaignsThe answer is complicated... We find a sizable benefit of #targeting (~70% more persuasive than showing everyone the best overall message) - but little additional benefit from *micro*targeting on numerous different characteristics simultaneouslyOur study used a 2-step process. First, we trained machine learning models on experimental data to learn which messages are most persuasive/for whom. Then, we recruited 5k new people randomized them into [Control; Random message; Best overall message; or Microtargeted message, chosen by training targeting model on 1, 2, or 4 different participant characteristics] & compared persuasion across conditions.Give the paper a read to see details!
(DIR) Post #AcXAMDUmyg2xHV6F5k by Drand@techhub.social
2023-12-06T13:20:23Z
7 likes, 7 repeats
@natalie Hi all, apologies for this, I didn't realize the bot was being poorly behaved - we've now stopped it. In terms of why we were doing the scraping, we are doing research on how content moderation policies vary across servers, and how this can help inform the Fediverse more broadly about effective approaches to content moderation. You can get more of a sense of the kind of research we do here: https://docs.google.com/document/d/1k2D4zVqkSHB1M9wpXtAe3UzbeE0RPpD_E2UpaPf6Lds/edit?usp=sharingSorry again about causing problems for folks! (And thanks to a couple of people for emailing me to let me know about this)
(DIR) Post #AcXFFRLFbD31y5E5Jo by Drand@techhub.social
2023-12-06T13:27:30Z
0 likes, 0 repeats
@stanford @natalie The email was much appreciated!! (Sorry again about the issue) https://techhub.social/@Drand/111533675192775824
(DIR) Post #Aca8sTQMZGCw7CP6Lw by Drand@techhub.social
2023-12-07T21:52:48Z
0 likes, 0 repeats
@atlas Thanks for your engagement! To clarify, the data we were collecting was on what news domains people on different servers share, and how toxic the language they use is etc; and then how that relates to the formal rules the server posts. But also we would LOVE to talk admins about how they actual think about / do content moderation etc (We have talked to some already, and are def interested in more!)
(DIR) Post #AihfoqmevBYVfJQRN2 by Drand@techhub.social
2024-02-07T16:30:45Z
0 likes, 0 repeats
🚨New WP🚨Field experiments with 33 million FB users & 75k Twitter users: Ads prompting users to think about accuracy reduce misinformation sharing! Accuracy prompts offer platforms a content-neutral approach that is scalable and preservers user autonomy https://osf.io/preprints/psyarxiv/u8anb
(DIR) Post #AihfpfNwmp5kbRFvQu by Drand@techhub.social
2024-03-25T16:52:38Z
0 likes, 0 repeats
🚨New PNASNexus🚨Reserach on misinformation & harmful language as often studied separately. Here we look at how they relate to each other, and find:-More harmful language in tweets w low-quality news links β=0.1 & in false headlines β=0.19 -Users who share more misinformation use more harmful language in non-news tweets β=0.13 PDF: https://academic.oup.com/pnasnexus/article/3/3/pgae111/7626151w @mohsenmosleh
(DIR) Post #Aihfs8BUgP2YV7bdlQ by Drand@techhub.social
2023-12-07T21:55:57Z
0 likes, 0 repeats
@ashten @natalie So you are not interested in a conversation re qualitative aspects of how you engage in content moderation?
(DIR) Post #Aihfs9Pi720SJVGVYu by Drand@techhub.social
2023-12-07T22:02:04Z
0 likes, 0 repeats
@ashten @natalie (My goal wasn't to dodge your questions, but rather to offer to answer them directly rather than in this very-prone-to-misunderstanding abbreviated format)
(DIR) Post #AihgG8zbKVGTxQUfS4 by Drand@techhub.social
2023-08-24T00:48:07Z
0 likes, 0 repeats
@NathanKalmoe what's the equivalent Stat for Dems?
(DIR) Post #AihgGB29jyUsHvhjiC by Drand@techhub.social
2023-08-24T01:35:20Z
0 likes, 0 repeats
@NathanKalmoe got it thanks! I bet a fair number of Dems would also agree that it's worth breaking rules to prevent eg another Trump term